Assessing the adsorption coefficient of diverse chemicals on polyethylene microplastics through a QSPR approach

Research on chemical adsorption onto microplastics is increasingly important in environmental studies. However, many existing models rely on basic structural properties, LSER descriptors, or 2D descriptor pools, often lacking robust external validation and applicability domain (AD) assessments. We revisited two literature studies to recompile a larger adsorption coefficient dataset (log K_d) of diverse organic chemicals onto polyethylene microplastics in pure water, applying rigorous data screening to ensure a higher-quality dataset and remove inconsistencies. log K_d values of 61 external chemicals were predicted from a dataset of 47 chemicals, including persistent, mobile and toxic (PMT) chemicals in both sets thereby extending predictions to PMT/vPvM compounds with limited experimental data. An extensive 3D descriptor set provided deeper mechanistic insights than 2D or physicochemical descriptors. Inclusion of 3D descriptors from dual-phase (gas and aqueous) geometry optimizations also improved mechanistic interpretation compared to gas-phase optimizations in literature. Our findings highlight the importance of strict external validation and a well-defined AD, using diverse validation metrics that improve upon the revisited studies. Adsorption correlated positively with lipophilicity and 3D-MoRSE descriptors, and negatively with -OH groups, ionization potential, and polarity. Model predictions aligned well with literature-reported log K_d values.